Benzodiazepine Anti-Anxiety Agents: Prevalence and Correlates of Use in a Southern Community
Total Page:16
File Type:pdf, Size:1020Kb
Benzodiazepine Anti-anxiety Agents: Prevalence and Correlates of Use in a Southern Community rn- rn Marvin Swartz, MD, Richard Landerman, PhD, Linda K George, PhD, Mary Lou Melville, MD, Dan Blazer, MD, PhD, and Karen Smith, PhD Introduction alence and patterns of benzodiazepine antianxiolytic drug use in the Piedmont re- Benzodiazepine anti-anxiety agents gion of North Carolina during 1982-83, uti- are the most widely prescribed psycho- lizing logistic regression analysis, which al- therapeutic drugs in the United States to- lows prediction of benzodiazepine use day.' First introduced in 1960, these drugs while introducing controls for potential rapidly achieved a lead position in the pre- confounding and mediating variables. scription drug market,2 stimulating public and professional debate over appropriate Method psychotropic drug use.3 Recent evidence, however, suggests that the prevalence and The present paper reports results patterns of psychotropic use, especially from Wave 1 of the Piedmont Health Sur- those of benzodiazepine anxiolytics, may vey, one site of the five-site National In- be changing and resulting in decreased stitute of Mental Health Epidemiologic use.4,5 The first detailed population survey Catchment Area program (NIMH- of psychotropic drug use, the National ECA).'1 The sampling frame for the Pied- Household Sample in 1970-71,6-9 found mont Health Survey (PHS) was a five- that 22 percent of American adults had county area in north central North used prescription psychotropic medica- Carolina, consisting of one urban county tion during the year 1969-70, with higher and four contiguous rural counties. The use among women and the elderly. The survey randomly sampled all housing most commonly used drug group, "minor units from segments throughout the catch- tranquilizer/daytime sedatives," was used ment area, using the Kish method,14 yield- by 8 percent of men and 20 percent of ing 3,798 interviews usable for this study; women. Use was greater in Whites and in further details of the study have been pub- divorced, separated, and widowed per- lished elsewhere.'5 In this paper analyses sons and the elderly.6 In 1979, the Na- are based on weighted data, taking into tional Survey of Psychotherapeutic Drug account household probability selection, Use documented, using weighted data, nonresponse, and the 1980 census demo- that an estimated total of 8 percent of men graphic proffle of adults in the five-county and 13.7 percent of women had used ben- catchment area. For significance tests, the zodiazepine anxiolytics in the previous data were downweighted to the original number of but year with the greatest use among respon- subjects, weighting adjust- ments remain. dents ages 5(04)42,10 Recent surveys of were obtained through use the St. Louis area population" and Great Diagnoses of the Diagnostic Interview Schedule Britain'2 found lower overall rates of use but similar sex and age usage profiles. Previous studies have relied primar- From the Psychiatric Epidemiology and Health ily on bivariate analyses when describing Services Research Program, Department of patterns of drug use, failing to take into Psychiatry, Duke University Medical Center. account the intercorrelations among pre- Address reprint requests to Marvin Swartz, dictors, including demographic, MD, Box 3173, Duke University Medical Cen- diagnos- ter, Durham, NC 27710. This paper, submitted tic, symptoms, and service use variables. to the Journal March 12, 1990, was revised and In the present study we examine the prev- accepted for publication December 12, 1990. May 1991, Vol. 81, No. 5 Prevalence and Correlates ofAnidolytic Use (DIS), a highly structured interview de- entered by stage. This format yields an signed for use by lay interviewers in epi- estimate of the effect of each variable net demiological studies and capable of gen- of all adjacent and prior variables and in- erating computer-based diagnoses for dicates whether the effects of prior varia- certain DSM-III disorders.'6 The diag- bles are mediated by subsequent varia- noses generated by the DIS were too nu- bles. The logistic regression procedure merous to include in a single regression calculates maximum likelihood estimates equation. Preliminary analyses ascer- for the parameters of a model that express tained that an affective disorders sum- the log odds ofan event (here, drug use) as mary measure (consisting of major de- a simple linear model.21"22 pressive disorders, bipolar disorders, dysthynmia, and bereavement), panic dis- Results order, and agoraphobia with panic were most strongly related to benzodiazepine Table 1 presents the demographic use. These measures were retained for profile for subjects participating in the subsequent analyses. The version of the survey.Table 2 focuses specifically on DIS used in Wave 1 of this survey and use of benzodiazepine anti-anxiety other ECA sites did not include one anx- agents-predominantly diazepam and iety disorders diagnosis, generalized anx- chlordiazepoxide-regardless of source. iety disorder, employed at certain ECA All other anxiolytics as well as all seda- sites in follow-up interviews. tives and hypnotics (benzodiazepine and In the present study psychotropic other) are excluded. Table 2 presents a drug use refers to any use within the year cross-tabular analysis of the weighted prior to interview, regardless ofamount or prevalence of benzodiazepine anxiolytic duration ofuse andwas elicitedbyspecific use among various demographic and drug use probes aided by photographic other groups of interest. For two contin- drug identification cards. Other measures uous variables, correlation coefficients used for this analysis include a 29-item are used: the correlation coefficient of psychic distress scale analogous to the drug use with number of negative life SCL-90,17 and the number of negative life events is .09 (p < .01) and that for drug events during the year prior to the inter- use with psychic distress symptoms is .20 view, measured by a scale of 20 major (p < .01). Benzodiazepine anxiolytic use events of the type proposed by Holmes is more prevalent among women, older and Rahe.18 Analyses of life events in the persons, Whites, and those with less ed- survey have been published else- ucation. The never married and married where.'9'20 Interviewers also recorded the have significantly lower prevalence of numberofoutpatientvisits to physical and use than the divorced and widowed. Re- mental health providers during the six spondents who have recently used either months prior to interview. The measure of physical or mental health care services outpatient health service utilization em- are much more likely to use these agents, ployed in this study is dichotomous, be- as are those with a DIS diagnosis of af- cause the use vs non-use measure has fective disorder, panic disorder, or ago- more explanatory power than a continu- raphobia with panic. ous (number of visits) measure. Demo- Table 3 presents results ofthe logistic due to higherlevelsofpsychic distress and graphic variables used in this report in- regression analyses. The "OR" coeffi- health service utilization among women. clude sex, age, race, education level, cient in the table is the antilogged regres- The race effect changes very little across marital status, and urban vs rural resi- sion coefficient and estimates how stages indicating no substantial effect of dence. Income is excluded because pre- changes in the independent variables mul- potential mediating variables. While the liminary analyses indicated that the rela- tiply the odds of using benzodiazepine coefficients for age fluctuate across stages, tionship between anxiolytic use and anxiolytics, holding other variables in the they remain largely unmediated and sub- income is weak. equation constant. Ninety-five percent stantial at stage VI. In stage II, increases The Appendix presents a six-stage confidence intervals are also given. The across each of four levels of education (as causal model predicting the dichotomous overall fit of the model is assessed by R indicated in Table 2) reduce the odds of measure of use vs nonuse of benzodi- statistic which is roughly analogous to the benzodiazepine use by .9. This initial ef- azepine anti-anxiety agents. A unidirec- multiple correlation coefficient in ordinary fect decreases in stages IV and V, indi- tional causal process is assumed with var- least squares regression after correcting cating that those with more education use iables in prior stages affecting those in for the number ofparameters estimated.22 benzodiazepines less, in part, because subsequent stages. No causal direction is In stage I, older persons and women they have fewer DIS diagnoses and psy- assumed for variables in the same stage. are both more likely than others to use chic distress symptoms. Compared to the In the multivariate logistic regression anal- benzodiazepines, while Blacks are less never married, the separated, divorced, ysis below, effects of the stage I variables likely. The gender effect decreases sub- and widowed are more likely to use ben- are estimated first, and the equation is stantially in stages V and VI indicating zodiazepine anxiolytics. These marital then expanded as groups of variables are that much of the initial sex difference is status effects are not mediated by varia- May 1991, Vol. 81, No. 5 American Journal of Public Health 593 Swartz et al. education, previously thought to be pre- dictors ofanxiolytic use, exert theireffects indirectlyvia theirrelationshipswith other